Cancer Diagnosis Classification Using Machine Learning Methods via Gene Expression Data

Detecting cancer cells that have diffused from their original cancer and got in the bloodstream, known as circulating tumor cells, shows a lot of promise as a prognostic indicator for individuals with cancer. Additionally, researchers are using non-invasive liquid biopsies to examine the genes expressed by circulating tumor cells in an effort to identify a potential cancer kind. In this work, we focused on predicting cancer by implementing machine learning techniques using the gene expression of circulating tumor cell data. The classifiers were initially trained to differentiate a specific cancer type from a mixture of other cancer types. Later, these models were modified to operate as multi-classifiers, enabling them to predict any of the seven cancer types, aiming to identify the bestperforming model and offering insights into the most effective approach for cancer prediction using gene expression data. Gradient Boosting, XGBoost, and AdaBoost algorithms are among the top performers across all cancer types due to their exceptional performance across all metrics, which comprise accuracy, precision, sensitivity, and F1 score values amidst 89 and 93% and AUC values of 92 and 96%.